Research On A Method Of Cloud Services Performace Ensurance Based On Component Replicas Add-delete

The business process oriented cloud service system, namely the software system based on component services in the cloud environment, occurred with the development of cloud computing and increasing service requirements. A cloud service provides services to users in the form of business flow, and each task in the flow is bound on the corresponding cloud component, so these component services work together to complete the whole cloud service. Every component service can have one or more replicas, deployed in the cloud environment resources pool. The component services of a cloud service have characteristics of the number changing of concurrent requests, the importance difference, the number difference of replicas and et al, which have influence on cloud service performance, such as performance bottlenecks, low VMs resource utility, insufficient numbers of component replicas. Therefore, it gradually becomes a research focus on performance ensurance mechanism for the cloud service based on mutiple replicas.On the basis of the previous service performance ensurance method, this thesis proposes a cloud service performance ensurance method based on component service replicas add-delete, referring to the service-oriented QoS incremental deployment method in service composition and analyzing the shortcomings. The performance ensurance method has two key studies, which are the cloud service bottleneck diagnosis method based on component service quality and importance, and the component service replicas add-delete method for cloud service performance ensurance. In the stage of bottleneck diagnosis, first of all, we do feature characterization and quantitative evalation for the factors affecting the running time-delay of a component service, which concludes the component service QoS. And then the invoking relationship among component services and their invoking frequences are analyzed to calculate the importance of each component service according to PageRank algorism. Finally, we combine the component service QoS with its importance to acquire the bottleneck component service, and the optimal component service. In the stage of replicas add-delete, we firstly determine to do replicas incremental deployment or deleting. And secondly, when doing replicas incremental deployment, it needs to analyze component services resource demands and VMs SLA violations, which can be integrated to judge the target VM. Replicas deleting requires analyzing resource usage of these VMs deployed by the optimal component services.Finally, we design simulation experiments to prove the accuracy of the bottleneck diagnosis method based on component services quality and importance, and the effectiveness of the component service replicas add-delete method for cloud service performance ensurance. Besides, the method is the preliminary study on the minimal resources to provide better services. Therefore, the cloud services performance ensurance method is an important reference for subsequent research.